# Axiom of choice - Proving an equivalence

I am reading a book on set theory and I got stuck on a simple step of a proof involving an equivalence of the axiom of choice.

Axiom of Choice (1) For every set $a\neq \emptyset$ there exists a function $f:\mathcal{P}\left (a\right )\setminus \left \{\emptyset\right \}\to a$ such that $f\left (b\right )\in b$ for every $b\subseteq a$, $b\neq \emptyset$

Axiom of Choice (2) Let $\left \{a_i\right \}_{i\in I}$ be a non-empty family of pairwise disjoint non-empty sets (i.e. $I\neq \emptyset$ and $a_i\cap a_j=\emptyset$ if $i\neq j$). Then there exists a set $d$ such that $d\cap a_i$ is a unit set for every $i\in I$.

I could understand the proof $\left (2\right )\Rightarrow \left (1\right )$, but I cannot understand the converse.

The proof starts by supposing that $\left (1\right )$ is true. We let $\left \{a_i\right \}_{i\in I}$ be a non-empty family of pairwise disjoint non-empty sets and define $a=\bigcup_{i\in I}a_i\neq \emptyset$. Therefore there exists a function $f:\mathcal{P}\left (a\right )\setminus \left \{\emptyset\right \}\to a$ be a function such that $f\left (b\right )\in b$ for every $b\in \mathcal{P}\left (a\right )\setminus \left \{\emptyset\right \}$. If $d$ is the image of $f$, then (the book says) we have $d\cap a_i=\left \{f\left (a_i\right )\right \}$. My question is: Why? It is obvious that one of the inclusions holds, but what about $\subseteq$?

• Restrict $f$ to $\{a_i\} \subset \mathcal{P}(a)$. – Zach Teitler Oct 14 '17 at 15:36
• @ZachTeitler: Eh? You don't need to use choice to choose from singletons. – Asaf Karagila Oct 14 '17 at 15:41
• That would be a restriction to a unit subset of the domain of $f$, the only thing I should do is applying $f$ to $a_i$ and then I obtain that $f\left (a_i\right )\in d\cap a_i$ which gives us the trivial inclusion. What about the other one? – solomeo paredes Oct 14 '17 at 15:43
• @AsafKaragila $a_i$ is not a singleton. Restrict $f$ to the collection $\{a_i\}_{i \in I}$. Version (1) provides a choice for every subset of $a$; we only care about the subsets $a_i$ for $i \in I$. – Zach Teitler Oct 14 '17 at 15:46
• @solomeoparedes Perhaps they defined $d$ in that way after restricting $f$ from all of $\mathcal{P}(a)$ to the subcollection $\{a_i\}_{i \in I}$? (Which by the way, some people would sometimes denote $\{a_i\}$, leaving the indexing $i \in I$ implicit, especially in certain contexts like writing $(a_n)$ for a sequence, or when they are typing on their phones.) – Zach Teitler Oct 14 '17 at 15:52

Recall that $a=\bigcup_{i\in I}a_i$ is trivially a superset of each of the $a_i$'s. Therefore each $a_i$ is an element of $\mathcal P(a)$.
By assumption, no $a_i$ is empty, so $a_i$ is in the domain of the choice function. Now take $\{f(a_i)\mid i\in I\}$ to be $d$, and by the fact each $a_i$ is disjoint we get that $d\cap a_i$ is a singleton for all $i\in I$.
• So, as I said on a previous comment: Should I assume that the book has an errata? Because the book defined $d$ as the whole image of $f$, not just the image of the $a_i$. – solomeo paredes Oct 14 '17 at 15:51